The most trusted bank in Europe held its premium for over 150 years on a single promise: it would never lend. It lent in secret. The premium was gone inside a year. The mechanism that made it trustworthy is the same mechanism that killed it.
A commitment is credible to the exact degree that it cannot be reversed. That is not a sentiment. It is a pricing rule, and it has a clean proof.
The Premium Is the Constraint
The Amsterdam Wisselbank, founded in 1609, held an agio for over 150 years: a standing premium of its bank money over the same weight of coin. Ask why anyone would pay more for a ledger entry than for the metal behind it. Because the bank was built to be incapable of debasing it. It held full reserve. It did not lend. Everyone knew it could not. The premium was the price of that constraint. The constraint was the product.
Then it lent. Quietly, to the Dutch East India Company and to the city of Amsterdam. The reserve was no longer whole. Nothing about the building changed. The clerks were the same, the vault was the same, the charter still said what it had always said. But the one promise that made the money worth more than metal was now false. When the secret surfaced in the 1780s, the agio collapsed. The bank was insolvent by 1790. A century and a half of premium, undone by the single act it existed to forbid.
Read it as a law. Trust is not produced by good intentions, and it is not produced by reputation. It is produced by the absence of an exit. Amotz Zahavi gave biology the identical result in 1975: a signal is believable only when it is too costly to fake. The peacock's tail is honest because it is a genuine handicap. The Wisselbank's premium was honest because the bank had genuinely tied its own hands. Remove the handicap and you remove the signal. There is no third option where the hands stay free and the promise stays believed.
Trust is not produced by good intentions, and it is not produced by reputation. It is produced by the absence of an exit.
The Second Clause of Conviction Capital
We have argued that conviction is the scarce asset under AI. Intelligence went abundant, so its price fell to zero, and the value moved to the one thing the machines do not supply: the willingness to commit resources under irreducible ambiguity. That is the case made in Conviction and the Pruning Problem, and it has a second clause we never stated. The Wisselbank states it for us.
A commitment earns the premium only by being genuinely impossible to back out of. Optionality and conviction are not two settings on a dial you can split the difference between. The moment a position can be quietly unwound, it stops paying the premium, because the market prices the exit and not the promise. This is why hedged conviction is a contradiction in terms. A board that commits while keeping the reversal warm has not committed. It has bought an expensive option and told itself it bought a position. The counterparty can see the difference even when the board cannot.
A Vault of Locked Doors Is a Tomb
Now the inversion, and it is exact. Run the same mechanism forward in time and across a whole book. Every credible commitment is, by construction, a door that cannot be reopened. Accumulate enough of them and you have not built a strategy. You have built a waqf.
The waqf is the Islamic endowment: an asset placed beyond sale, beyond seizure, beyond reform, in perpetuity, its uses fixed by the founder's hand. It financed hospitals and madrasas that outlived the dynasties that built them, precisely because no later ruler could touch it. That is the same irreversibility that bought the Wisselbank its premium, and it worked. Then it kept working. Timur Kuran's account of the long economic divergence is that a large share of Ottoman productive land sat frozen in perpetual endowment, unable to answer a single market signal, its purposes fixed at the founder's death and binding on a world the founder never saw. The mechanism that made each commitment credible made the whole economy unable to move.
This is the pruning problem arriving by a new road. We called the un-prunable portfolio a landfill and blamed mismanagement, zombie probes nobody had the nerve to kill. The waqf says it is worse than that. A book of maximally credible commitments becomes a landfill by construction, not by cowardice. The firms with the strongest conviction are the most exposed, because their commitments are the hardest to reverse, which was the entire point of making them.
Both pillars sit on this fault line. In market structure, a central counterparty's default waterfall and a clearing fund's rules are the locked plumbing that makes a market safe to trade. That plumbing is trustworthy for the Wisselbank reason: it cannot be changed under pressure, so no member can lobby it loose in a crisis. And it ossifies for the waqf reason. The rule written to guarantee safety in one regime is the rule that cannot bend when a regime it was never written for arrives. Irreversible safety and irreversible rigidity are the same property seen from two distances.
In professional services, McKinsey now says it underwrites the outcomes of roughly a third of its revenue. That is conviction made balance-sheet real, a book of positions the firm can no longer advise its way out of. It is the correct move. It is also a waqf forming in real time. A book of underwritten outcomes is a book of doors that will not reopen, and the better the firm is at conviction, the faster that book sets solid.
A book of maximally credible commitments becomes a landfill by construction, not by cowardice.
The Answer Is a Half-Life, Not a Level
So: commit, or stay flexible? Wrong question. The English common law answered it three centuries ago, and the answer is neither. The Rule Against Perpetuities was settled in the Duke of Norfolk's Case in 1682. A property interest must vest, if at all, within a life in being plus twenty-one years. The dead hand may reach into the future, but only so far. The rule lets the founder bind what comes next, which preserves the credibility, and then forces the binding to lapse, which preserves the power to reallocate. It is a designed sunset on irreversibility. It priced the exact problem the waqf could not solve, and it priced it as a horizon rather than as a permission.
That is the design rule conviction has been missing. Not weaker commitment. A commitment half-life. A conviction must be irreversible enough to be credible and prunable enough to be survivable. The only way to hold both at once is to date-stamp the irreversibility at the moment you make it. Commit hard, in public, with a real cost to reversal, and with a vesting horizon written into the commitment itself, after which it dies unless it is deliberately renewed against fresh signal. The renewal is the prune. The expiry is what keeps the book from setting solid while you are still proud of it.
A sunset weakens the signal, the cynic will say, since everyone knows the door reopens in the end. It does not, because a life plus twenty-one years is longer than any competitor's patience and longer than almost every strategic window that matters. The commitment is dead before the exit ever becomes relevant.
Conviction without an expiry is a vault you have locked from the inside, and the premium you earned for locking it is the premium the next generation will pay to be let out.